Skip to content
Snippets Groups Projects
test_jfa.py 11.38 KiB
#!/usr/bin/env python
# vim: set fileencoding=utf-8 :
# Laurent El Shafey <Laurent.El-Shafey@idiap.ch>
# Wed Feb 15 23:24:35 2012 +0200
#
# Copyright (C) 2011-2013 Idiap Research Institute, Martigny, Switzerland

"""Tests on the JFA-based machines
"""

import os, sys
import unittest
import math
import bob
import numpy, numpy.linalg
import tempfile

def estimate_x(dim_c, dim_d, mean, sigma, U, N, F):
  # Compute helper values
  UtSigmaInv = {}
  UtSigmaInvU = {}
  dim_ru = U.shape[1]
  for c in range(dim_c):
    start                       = c*dim_d
    end                         = (c+1)*dim_d
    Uc                          = U[start:end,:]
    UtSigmaInv[c]  = Uc.transpose() / sigma[start:end]
    UtSigmaInvU[c] = numpy.dot(UtSigmaInv[c], Uc);

  # I + (U^{T} \Sigma^-1 N U)
  I_UtSigmaInvNU = numpy.eye(dim_ru, dtype=numpy.float64) 
  for c in range(dim_c):
    I_UtSigmaInvNU = I_UtSigmaInvNU + UtSigmaInvU[c] * N[c]
  
  # U^{T} \Sigma^-1 F
  UtSigmaInv_Fnorm = numpy.zeros((dim_ru,), numpy.float64)
  for c in range(dim_c):
    start             = c*dim_d
    end               = (c+1)*dim_d
    Fnorm             = F[c,:] - N[c] * mean[start:end]
    UtSigmaInv_Fnorm  = UtSigmaInv_Fnorm + numpy.dot(UtSigmaInv[c], Fnorm)
 
  return numpy.linalg.solve(I_UtSigmaInvNU, UtSigmaInv_Fnorm)

def estimate_ux(dim_c, dim_d, mean, sigma, U, N, F):
  return numpy.dot(U, estimate_x(dim_c, dim_d, mean, sigma, U, N, F))


class FATest(unittest.TestCase):
  """Performs various FA tests."""

  def test01_JFABase(self):

    # Creates a UBM
    weights = numpy.array([0.4, 0.6], 'float64')
    means = numpy.array([[1, 6, 2], [4, 3, 2]], 'float64')
    variances = numpy.array([[1, 2, 1], [2, 1, 2]], 'float64')
    ubm = bob.machine.GMMMachine(2,3)
    ubm.weights = weights
    ubm.means = means
    ubm.variances = variances

    # Creates a JFABase
    U = numpy.array([[1, 2], [3, 4], [5, 6], [7, 8], [9, 10], [11, 12]], 'float64')
    V = numpy.array([[6, 5], [4, 3], [2, 1], [1, 2], [3, 4], [5, 6]], 'float64')
    d = numpy.array([0, 1, 0, 1, 0, 1], 'float64')
    m = bob.machine.JFABase(ubm)
    self.assertTrue( m.dim_ru == 1)
    self.assertTrue( m.dim_rv == 1)

    # Checks for correctness
    m.resize(2,2)
    m.u = U
    m.v = V
    m.d = d
    self.assertTrue( (m.u == U).all() )
    self.assertTrue( (m.v == V).all() )
    self.assertTrue( (m.d == d).all() )
    self.assertTrue( m.dim_c == 2)
    self.assertTrue( m.dim_d == 3)
    self.assertTrue( m.dim_cd == 6)
    self.assertTrue( m.dim_ru == 2)
    self.assertTrue( m.dim_rv == 2)

    # Saves and loads
    filename = str(tempfile.mkstemp(".hdf5")[1])
    m.save(bob.io.HDF5File(filename, 'w'))
    m_loaded = bob.machine.JFABase(bob.io.HDF5File(filename))
    m_loaded.ubm = ubm
    self.assertTrue( m == m_loaded )
    self.assertFalse( m != m_loaded )
    self.assertTrue( m.is_similar_to(m_loaded) )

    # Copy constructor
    mc = bob.machine.JFABase(m)
    self.assertTrue( m == mc )

    # Variant
    mv = bob.machine.JFABase()
    # Checks for correctness
    mv.ubm = ubm
    mv.resize(2,2)
    mv.u = U
    mv.v = V
    mv.d = d
    self.assertTrue( (m.u == U).all() )
    self.assertTrue( (m.v == V).all() )
    self.assertTrue( (m.d == d).all() )
    self.assertTrue( m.dim_c == 2)
    self.assertTrue( m.dim_d == 3)
    self.assertTrue( m.dim_cd == 6)
    self.assertTrue( m.dim_ru == 2)
    self.assertTrue( m.dim_rv == 2)

    # Clean-up
    os.unlink(filename)

  def test02_ISVBase(self):

    # Creates a UBM
    weights = numpy.array([0.4, 0.6], 'float64')
    means = numpy.array([[1, 6, 2], [4, 3, 2]], 'float64')
    variances = numpy.array([[1, 2, 1], [2, 1, 2]], 'float64')
    ubm = bob.machine.GMMMachine(2,3)
    ubm.weights = weights
    ubm.means = means
    ubm.variances = variances

    # Creates a ISVBase
    U = numpy.array([[1, 2], [3, 4], [5, 6], [7, 8], [9, 10], [11, 12]], 'float64')
    d = numpy.array([0, 1, 0, 1, 0, 1], 'float64')
    m = bob.machine.ISVBase(ubm)
    self.assertTrue( m.dim_ru == 1)

    # Checks for correctness
    m.resize(2)
    m.u = U
    m.d = d
    self.assertTrue( (m.u == U).all() )
    self.assertTrue( (m.d == d).all() )
    self.assertTrue( m.dim_c == 2)
    self.assertTrue( m.dim_d == 3)
    self.assertTrue( m.dim_cd == 6)
    self.assertTrue( m.dim_ru == 2)

    # Saves and loads
    filename = str(tempfile.mkstemp(".hdf5")[1])
    m.save(bob.io.HDF5File(filename, 'w'))
    m_loaded = bob.machine.ISVBase(bob.io.HDF5File(filename))
    m_loaded.ubm = ubm
    self.assertTrue( m == m_loaded )
    self.assertFalse( m != m_loaded )
    self.assertTrue( m.is_similar_to(m_loaded) )

    # Copy constructor
    mc = bob.machine.ISVBase(m)
    self.assertTrue( m == mc )

    # Variant
    mv = bob.machine.ISVBase()
    # Checks for correctness
    mv.ubm = ubm
    mv.resize(2)
    mv.u = U
    mv.d = d
    self.assertTrue( (m.u == U).all() )
    self.assertTrue( (m.d == d).all() )
    self.assertTrue( m.dim_c == 2)
    self.assertTrue( m.dim_d == 3)
    self.assertTrue( m.dim_cd == 6)
    self.assertTrue( m.dim_ru == 2)

    # Clean-up
    os.unlink(filename)
  
  def test03_JFAMachine(self):

    # Creates a UBM
    weights = numpy.array([0.4, 0.6], 'float64')
    means = numpy.array([[1, 6, 2], [4, 3, 2]], 'float64')
    variances = numpy.array([[1, 2, 1], [2, 1, 2]], 'float64')
    ubm = bob.machine.GMMMachine(2,3)
    ubm.weights = weights
    ubm.means = means
    ubm.variances = variances

    # Creates a JFABase
    U = numpy.array([[1, 2], [3, 4], [5, 6], [7, 8], [9, 10], [11, 12]], 'float64')
    V = numpy.array([[6, 5], [4, 3], [2, 1], [1, 2], [3, 4], [5, 6]], 'float64')
    d = numpy.array([0, 1, 0, 1, 0, 1], 'float64')
    base = bob.machine.JFABase(ubm,2,2)
    base.u = U
    base.v = V
    base.d = d

    # Creates a JFAMachine
    y = numpy.array([1,2], 'float64')
    z = numpy.array([3,4,1,2,0,1], 'float64')
    m = bob.machine.JFAMachine(base)
    m.y = y
    m.z = z
    self.assertTrue( m.dim_c == 2)
    self.assertTrue( m.dim_d == 3)
    self.assertTrue( m.dim_cd == 6)
    self.assertTrue( m.dim_ru == 2)
    self.assertTrue( m.dim_rv == 2)
    self.assertTrue( (m.y == y).all() )
    self.assertTrue( (m.z == z).all() )
    
    # Saves and loads
    filename = str(tempfile.mkstemp(".hdf5")[1])
    m.save(bob.io.HDF5File(filename, 'w'))
    m_loaded = bob.machine.JFAMachine(bob.io.HDF5File(filename))
    m_loaded.jfa_base = base
    self.assertTrue( m == m_loaded )
    self.assertFalse( m != m_loaded )
    self.assertTrue(m.is_similar_to(m_loaded))

    # Copy constructor
    mc = bob.machine.JFAMachine(m)
    self.assertTrue( m == mc )

    # Variant
    mv = bob.machine.JFAMachine()
    # Checks for correctness
    mv.jfa_base = base
    m.y = y
    m.z = z
    self.assertTrue( m.dim_c == 2)
    self.assertTrue( m.dim_d == 3)
    self.assertTrue( m.dim_cd == 6)
    self.assertTrue( m.dim_ru == 2)
    self.assertTrue( m.dim_rv == 2)
    self.assertTrue( (m.y == y).all() )
    self.assertTrue( (m.z == z).all() )

    # Defines GMMStats
    gs = bob.machine.GMMStats(2,3)
    log_likelihood = -3.
    T = 1
    n = numpy.array([0.4, 0.6], 'float64')
    sumpx = numpy.array([[1., 2., 3.], [4., 5., 6.]], 'float64')
    sumpxx = numpy.array([[10., 20., 30.], [40., 50., 60.]], 'float64')
    gs.log_likelihood = log_likelihood
    gs.t = T
    gs.n = n
    gs.sum_px = sumpx
    gs.sum_pxx = sumpxx

    # Forward GMMStats and check estimated value of the x speaker factor
    eps = 1e-10
    x_ref = numpy.array([0.291042849767692, 0.310273618998444], 'float64')
    score_ref = -2.111577181208289
    score = m.forward(gs)
    self.assertTrue( numpy.allclose(m.__x__, x_ref, eps) )
    self.assertTrue( abs(score_ref-score) < eps )

    # x and Ux
    x = numpy.ndarray((2,), numpy.float64)
    m.estimate_x(gs, x)
    x_py = estimate_x(m.dim_c, m.dim_d, ubm.mean_supervector, ubm.variance_supervector, U, n, sumpx)
    self.assertTrue( numpy.allclose(x, x_py, eps))

    ux = numpy.ndarray((6,), numpy.float64)
    m.estimate_ux(gs, ux)
    ux_py = estimate_ux(m.dim_c, m.dim_d, ubm.mean_supervector, ubm.variance_supervector, U, n, sumpx)
    self.assertTrue( numpy.allclose(ux, ux_py, eps))
    self.assertTrue( numpy.allclose(m.__x__, x, eps) )

    score = m.forward_ux(gs, ux)
    self.assertTrue( abs(score_ref-score) < eps )

    # Clean-up
    os.unlink(filename)
  
  def test04_ISVMachine(self):

    # Creates a UBM
    weights = numpy.array([0.4, 0.6], 'float64')
    means = numpy.array([[1, 6, 2], [4, 3, 2]], 'float64')
    variances = numpy.array([[1, 2, 1], [2, 1, 2]], 'float64')
    ubm = bob.machine.GMMMachine(2,3)
    ubm.weights = weights
    ubm.means = means
    ubm.variances = variances

    # Creates a JFABaseMachine
    U = numpy.array([[1, 2], [3, 4], [5, 6], [7, 8], [9, 10], [11, 12]], 'float64')
    V = numpy.array([[0], [0], [0], [0], [0], [0]], 'float64')
    d = numpy.array([0, 1, 0, 1, 0, 1], 'float64')
    base = bob.machine.ISVBase(ubm,2)
    base.u = U
    base.v = V
    base.d = d

    # Creates a JFAMachine
    z = numpy.array([3,4,1,2,0,1], 'float64')
    m = bob.machine.ISVMachine(base)
    m.z = z
    self.assertTrue( m.dim_c == 2)
    self.assertTrue( m.dim_d == 3)
    self.assertTrue( m.dim_cd == 6)
    self.assertTrue( m.dim_ru == 2)
    self.assertTrue( (m.z == z).all() )

    # Saves and loads
    filename = str(tempfile.mkstemp(".hdf5")[1])
    m.save(bob.io.HDF5File(filename, 'w'))
    m_loaded = bob.machine.ISVMachine(bob.io.HDF5File(filename))
    m_loaded.isv_base = base
    self.assertTrue( m == m_loaded )
    self.assertFalse( m != m_loaded )
    self.assertTrue(m.is_similar_to(m_loaded))

    # Copy constructor
    mc = bob.machine.ISVMachine(m)
    self.assertTrue( m == mc )

    # Variant
    mv = bob.machine.ISVMachine()
    # Checks for correctness
    mv.isv_base = base
    m.z = z
    self.assertTrue( m.dim_c == 2)
    self.assertTrue( m.dim_d == 3)
    self.assertTrue( m.dim_cd == 6)
    self.assertTrue( m.dim_ru == 2)
    self.assertTrue( (m.z == z).all() )

    # Defines GMMStats
    gs = bob.machine.GMMStats(2,3)
    log_likelihood = -3.
    T = 1
    n = numpy.array([0.4, 0.6], 'float64')
    sumpx = numpy.array([[1., 2., 3.], [4., 5., 6.]], 'float64')
    sumpxx = numpy.array([[10., 20., 30.], [40., 50., 60.]], 'float64')
    gs.log_likelihood = log_likelihood
    gs.t = T
    gs.n = n
    gs.sum_px = sumpx
    gs.sum_pxx = sumpxx

    # Forward GMMStats and check estimated value of the x speaker factor
    eps = 1e-10
    x_ref = numpy.array([0.291042849767692, 0.310273618998444], 'float64')
    score_ref = -3.280498193082100

    score = m.forward(gs)
    self.assertTrue( numpy.allclose(m.__x__, x_ref, eps) )
    self.assertTrue( abs(score_ref-score) < eps )

    # Check using alternate forward() method
    Ux = numpy.ndarray(shape=(m.dim_cd,), dtype=numpy.float64)
    m.estimate_ux(gs, Ux)
    score = m.forward_ux(gs, Ux)
    self.assertTrue( abs(score_ref-score) < eps )

    # x and Ux
    x = numpy.ndarray((2,), numpy.float64)
    m.estimate_x(gs, x)
    x_py = estimate_x(m.dim_c, m.dim_d, ubm.mean_supervector, ubm.variance_supervector, U, n, sumpx)
    self.assertTrue( numpy.allclose(x, x_py, eps))

    ux = numpy.ndarray((6,), numpy.float64)
    m.estimate_ux(gs, ux)
    ux_py = estimate_ux(m.dim_c, m.dim_d, ubm.mean_supervector, ubm.variance_supervector, U, n, sumpx)
    self.assertTrue( numpy.allclose(ux, ux_py, eps))
    self.assertTrue( numpy.allclose(m.__x__, x, eps) )

    score = m.forward_ux(gs, ux)
    self.assertTrue( abs(score_ref-score) < eps )

    # Clean-up
    os.unlink(filename)